Why real-time scheduling theory still matters

Sanjoy Baruah

The University of North Carolina at Chapel Hill

Our discipline

Has over-emphasized the theory

is about systems …that require formal/ theoretical analysis

- A distinguishing characteristic of the discipline

Is starting to remedy this

Our discipline = Systems + Theory

-The “only theory?” test -Hard to get a uniprocessor paper into ECRTS/ RTSS!-Special issues/ invited talks are systems-oriented

Let’s not over-compensate

Scheduling theory remains relevant to real-time systems.

Scheduling theory remains relevant to real-time systems. Some forms of scheduling theory are particularly important.

Scheduling theory remains relevant to real-time systems. Some forms of scheduling theory are particularly important. These important forms can be identified.

Scheduling theory remains relevant to real-time systems. Some forms of scheduling theory are particularly important. These important forms can be identified.

Outline of presentation

1. Why theory mattered (and matters)2. What forms of scheduling theory are important3. What specific areas are most important

Thesis

why theory? – what kinds? – which areas?why theory? – what kinds? – which areas?why theory? – what kinds? – which areas?

Our discipline = Systems + Theory

The window of scarcity

Real-time systems often have resource constraints

hardware resources

systemspecifications

impossible toimplement

window of scarcity

easy to implement

Why theory?

why theory? – what kinds? – which areas?

Why theory?

Our discipline = Systems + Theory

The window of scarcity

Often safety-critical

• Very high cost of error

• Must be validated correct by Certification Authorities

Use theory to prove correctness

why theory? – what kinds? – which areas?

Real-time systems often have resource constraints

Why theory?

The window of scarcity

Often safety-critical

why theory? – what kinds? – which areas?

What counts as good theory?

Should potentially be useful

The “pure” sciences legitimately take the discovery of facts and laws as an end in itself. A new fact, a new law is an accomplishment, worthy of publication. But in [computer science and engineering] novelty in itself has no merit.. We test our artifacts by their usefulness and their costs, not their novelty.

- Fred Brooks. The computer scientist as toolsmith. (CACM 1996)

2. need not be useful immediately

1. new facts and laws are needed

- but present them elsewhere (e.g., theory conferences)

Caveats

why theory? – what kinds? – which areas?

“Industrially relevant research is … research that is usable when industry decides it is relevant”

-Bjorn Andersson

What counts as good theory?

Should potentially be useful

At the appropriate level of abstraction

* Should be applicable to an entire class of systems

- else, just a project report* Therefore, on abstract task/ machine models

- highlights a few salient features or principles

- E.g., for hard-real-time schedulability analysis

* Using the appropriate abstractions

In that empire, the art of cartography attained such perfection that […] the cartographers guilds struck a map of the empire whose size was that of the empire, and which coincided point for point with it.

In the Deserts of the West, still today, there are Tattered Ruins of that Map, inhabited by Animals and Beggars.

-Jorge Luis Borges

In that empire, the art of cartography attained such perfection that […] the cartographers guilds struck a map of the empire whose size was that of the empire, and which coincided point for point with it.

In the deserts of the west, still today, there are tattered ruins of that map, inhabited by animals and beggars.

-Jorge Luis Borges

Obtaining appropriate abstractions is an important research area

* Computationally tractable abstractions

why theory? – what kinds? – which areas?

Promising research areas

Multiprocessors

Component-based design

Mixed criticalities

why theory? – what kinds? – which areas?

Promising research areas

Multiprocessors

Component-based design

* Future RT systems will be multiprocessor ones -the multicore revolution

* Multiprocessor scheduling theory is not mature enough

* Important questions- models- metrics- scheduling algorithms

Mixed criticalities

Big-picture question: what critical insights are needed for multiproc. scheduling?

why theory? – what kinds? – which areas?

Promising research areas

Multiprocessors

Component-based design

Open systems and componentization

- more powerful platforms

- SWaP considerations

- software engineering issues

Some work has been done…but much remains

- abstraction and interface specification- an algebra for composition

Mixed criticalities

why theory? – what kinds? – which areas?

Size, Weight, and Power (i.e., Energy)

Promising research areas

Multiprocessors

Component-based design

Mixed criticalities: Promising research areas

Mixed criticalities

why theory? – what kinds? – which areas?

Different sub-systems have different certification requirements- Defense avionics example.

Some sub-systems are more important than others - Automotive example: ABS vs car stereo

- flight-critical certification: cycle-counting under pessimistic assumptions- mission-critical validation: extensive experimentation

Example: Determining worst-case execution time (WCET)

Flight critical: certified by Certification Authorities

Mission-critical: validated by design team

and mission-critical functionalitiesFlight-critical

Mixed criticalities: Promising research areas

why theory? – what kinds? – which areas?

J1 is flight-critical; J2 is mission-criticalBoth arrive at t=0; have deadlines at t=10 WCET of J1 is 6; WCET of J2 is 5

0 2 4 6 8 10time

6 + 5 > 10 not schedulable

But…- flight-criticality certification does not need J2 to meet its deadline- for mission-critical validation, J1’s WCET of 6 may be too pessimistic

* Suppose J1’s WCET, obtained by extensive experimentation, is 5

Priority-based scheduling: J1 > J2

Mixed criticalities: Promising research areas

why theory? – what kinds? – which areas?

0 2 4 6 8 10time

* Suppose J1’s WCET, obtained by extensive experimentation, is 5

Flight-criticality certification

J2 misses deadlineJ1 meets deadline

But…- flight-criticality certification does not need J2 to meet its deadline- for mission-critical validation, J1’s WCET of 6 may be too pessimistic

J1 is flight-critical; J2 is mission-criticalBoth arrive at t=0; have deadlines at t=10 WCET of J1 is 6; WCET of J2 is 5

Priority-based scheduling: J1 > J2

Mixed criticalities: Promising research areas

why theory? – what kinds? – which areas?

0 2 4 6 8 10time

But…- flight-criticality certification does not need J2 to meet its deadline- for mission-critical validation, J1’s WCET of 6 may be too pessimistic

* Suppose J1’s WCET, obtained by extensive experimentation, is 5

Mission-critical validation

J1 meets deadline

J2 meets deadline

J1 is flight-critical; J2 is mission-criticalBoth arrive at t=0; have deadlines at t=10 WCET of J1 is 6; WCET of J2 is 5

Validated at both criticalities

Priority-based scheduling: J1 > J2

Mixed criticalities: Promising research areas

why theory? – what kinds? – which areas?

Validated at both criticalitiesJ1 is flight-critical; J2 is mission-criticalBoth arrive at t=0; have deadlines at t=10 WCET of J1 is 6; WCET of J2 is 5

The same system is being validated, twice

Flight-critical certification Mission-critical validation

of only a subset of the system

at a very high level of assurance

of the entire system

at a lower level of assurance

Mixed criticalities: Promising research areas

Interesting open issues:- How do we represent MC systems?- How do we reason about them?

- “parallel worlds”? space-time partitioning?- What scheduling strategies are suitable?

why theory? – what kinds? – which areas?

Summary

Scheduling theory remains relevant to real-time systems.

A thesis…

…and its justification

• The window of scarcity; safety-critical nature of applications

Summary

Scheduling theory remains relevant to real-time systems. Some forms of scheduling theory are particularly important.

A thesis…

…and its justification

• The window of scarcity; safety-critical nature of applications

• Must be relevant; must be at appropriate level of abstraction

Scheduling theory remains relevant to real-time systems. Some forms of scheduling theory are particularly important. These important forms can be identified.

Summary

A thesis…

…and its justification

• The window of scarcity; safety-critical nature of applications

• Must be relevant; must be at appropriate level of abstraction

• Multiprocessors; component-based design; and mixed

criticalities

SCHEDULING THEORY

REAL-TIME SYSTEMS

Scheduling theory remains relevant to real-time systems. Some forms of scheduling theory are particularly important. These important forms can be identified.

Summary

A thesis…

…and its justification

• The window of scarcity; safety-critical nature of applications

• Must be relevant; must be at appropriate level of abstraction

• Multiprocessors; component-based design; and mixed

criticalities